IntroducIng the dell Perc 6 FamIly oF SaS raId … · By Bhanu Prakash Dixit Sanjay Tiwari ......

Storage

DELL POWER SOLUTIONS | May 200888 Reprinted from Dell Power Solutions, May 2008. Copyright © 2008 Dell Inc. All rights reserved.

As enterprise data requirements continue

to increase, deploying powerful, easy-to-

manage RAID solutions can become essen-

tial for organizations seeking to simplify their IT

environments while increasing controller and stor-

age reliability and server uptime. The Dell PowerEdge

Expandable RAID Controller (PERC) line of storage

controllers has evolved to address these data stor-

age requirements, providing innovative features to

help support complex server-based storage environ-

ments, increase server uptime, and reduce adminis-

trative burdens.

The PERC 6 family of Serial Attached SCSI (SAS)

RAID controllers supports SAS devices and Dell-

qualified Serial ATA (SATA) devices. It consists of

three PCI Express (PCIe)–based, 3 Gbps SAS RAID

controllers: the PERC 6/E adapter (which fits in a

standard x8 PCIe slot supporting Dell PowerVault™

direct attach storage), the PERC 6/i adapter

(which fits in a standard x8 PCIe slot supporting

internal storage for Dell PowerEdge servers and

Dell Precision™ workstations), and the PERC 6/i inte-

grated card (which supports internal storage for

Dell PowerEdge servers).

Each PERC 6 controller offers a 256 MB double

data rate 2 (DDR2) cache memory, with the

PERC 6/E supporting upgrades to 512 MB to allow

quick access to additional data and help increase

performance. PERC 6 controllers also include a bat-

tery backup unit to help maintain data consistency

and protect data in the controller cache during a

system power and/or controller failure. PERC 6 con-

trollers support RAID levels 0, 1, 5, 6, 10, 50, and 60.

The evolution of the PERC 6 from the PERC 5

reflects the Dell focus on simplifying IT, advancing

green technology, and designing total data solutions.

PERC 6 controllers take advantage of the PCIe archi-

tecture and include an enhanced controller chip

design to help substantially increase throughput and

remove the controller as a performance bottleneck

(see Figure 1). The optimization of the RAID design,

which incorporates the I/O processor and I/O control-

ler on a single RAID-on-a-chip (ROC) solution, enables

a significant reduction in power consumption for

PERC 6 controllers compared with equivalent

PERC 5 controllers. In addition, using a true PCIe solu-

tion by removing the PCI Extended (PCI-X) bridge

helps significantly increase performance under

sequential workloads. Figure 2 provides a compre-

hensive comparison of PERC 5 and PERC 6 features

and specifications.

PERC 6 controllers are designed to provide

enhanced performance, simplified management,

increased reliability and fault tolerance, optimized

The Dell™ PowerEdge™ Expandable RAID Controller (PERC) 6 family of enterprise-class Serial Attached SCSI (SAS) RAID controllers is designed for enhanced performance, increased reliability and fault tolerance, and simplified management—providing a powerful, easy-to-manage way to create a robust infrastructure and help maximize server uptime.

By Bhanu Prakash Dixit

Sanjay Tiwari

Kedar Vaze

Joe H. Trickey III

IntroducIng the dell Perc 6 FamIly oF SaS raId controllerS

Related Categories:

Dell PowerEdge RAID Controller (PERC)

PERC 6

RAID

Serial Attached SCSI (SAS)

Storage

Visit DELL.COM/PowerSolutions

for the complete category index.

http://DELL.COM/PowerSolutions

89DELL.COM/PowerSolutionsReprinted from Dell Power Solutions, May 2008. Copyright © 2008 Dell Inc. All rights reserved.

deployment with Dell PowerVault storage,

and flexible RAID configuration and man-

agement tools. The controllers support

Microsoft® Windows Server® 2003 with

Service Pack 1 (SP1) (32-bit and 64-bit),

Microsoft Windows Server 2008 (32-bit

and 64-bit), Red Hat® Enterprise Linux® 4

Update 5 and Red Hat Enterprise Linux 5

(32-bit and 64-bit), Novell® SUSE® Linux

Enterprise Server 10 (64-bit), Microsoft

Windows® XP with SP2 (32-bit and 64-bit),

and Microsoft Windows Vista® (32-bit and

64-bit) operating systems.

enhanced performanceThe PERC 6 is designed to increase per-

formance in each of the four key areas

that can affect controller performance:

hardware features, caching policies, read

and I/O policies, and software features.

Each is important and should be consid-

ered when evaluating a RAID controller in

a host server along with other server fea-

tures that may affect performance, such

as the OS and hard drives.

Organizations should also keep in

mind that different applications display

different workload characteristics, and

that these workload footprints can affect

server performance. However, many

applications can be categorized as having

a sequential read/write profile or a

random read/write profile. Sequential

read/write applications typically include

media streaming, backup and restore,

and network attached storage and near-

line storage applications. Workstation

and file servers also handle large data file

read/write requests, making sequential

data performance a critical performance

requirement in enterprise environments.

For these applications, storage subsys-

tem performance is typically measured

by sequential throughput, in megabytes

per second.

Random read/write applications typi-

cally include transaction processing

applications such as databases, Web-

based e-commerce applications, and

other multitasking business solutions. For

these applications, storage subsystem

performance is typically measured in

I/Os per second.

hardware featuresThe PERC 6 continues the Dell focus on

designing RAID solutions based on the

latest interface technologies, from PCI to

PCIe, from parallel to serial. Employing a

ROC integrated circuit (IC) featuring a

high-speed, 64-bit, error-correcting code

(ECC)–protected DDR2 SDRAM interface;

a hardware RAID assist engine for parity

Bottleneck

Intel®

IOP333processor

DDR2-400 LSISAS1068controller IC

x8 PCIe

x83 Gbps

SAS

3.2 GB/secpeak

throughput

1 GB/secpeak

throughput

2.4 GB/secpeak throughput

PERC 5 architecture

Bottleneck

CPUcomplexDDR2-667 SAS

core

x8 PCIe2 GB/sec

peak throughput

x83 Gbps

SAS

Over 4 GB/secpeak

throughput

Over 4 GB/secpeak

throughput

2.4 GB/secpeak throughput

PERC 6 architecture

Hostchipset

CPU CPU

LSI 1078ROC

2 GB/secpeak

throughput

PCI-X

Figure 1. Dell PERC 5 and PERC 6 controller architectures

perc 5 perc 6

RAID levels 0, 1, 5, 10, and 50 0, 1, 5, 6, 10, 50, and 60

Enclosures per port Up to three Up to three

Ports 2 x 4 wide ports 2 x 4 wide ports

Write policies Write-through and write-back Write-through and write-back

Read policies Normal, read-ahead, and adaptive Normal, read-ahead, and adaptive

Virtual disks per controller Up to 64 Up to 64

Cache memory size 256 MB 256 MB (up to 512 MB for PERC 6/E)

PCIe link width x8 x8

256 KB, 512 KB, and 1,024 KB stripe sizes 4

SATA NCQ support 4

Global, dedicated, and affinity hot spares

Global and dedicated hot spares only 4

Online capacity expansion 4 4

Hot-swappable drives 4 4

Mixed-capacity physical disk support 4 4

Hardware exclusive OR (XOR) assistance 4 4

Figure 2. Feature comparison of Dell PERC 5 and PERC 6 controllers

Storage


calculations; and support for an Intelligent

I/O (I2O) message unit helps maximize

performance. The controller is powered

by a state-of-the-art LSI 1078 ROC con-

troller, which includes both instruction

and data cache to help provide an addi-

tional performance boost. This solid hard-

ware foundation is designed to simplify

integration of new RAID firmware features

and performance increases in future

release updates.

caching policiesFlexible caching policies in PERC 6 control-

lers allow administrators to tune cache writ-

ing schemes to help optimize performance

or maximize data protection, including

choosing between write-through caching

and write-back caching. With write-through

caching enabled, the controller transfers a

block of data directly to the disk, and does

not send an acknowledgment of data trans-

fer completion to the host system until the

data is committed to the disk subsystem.

With write-back caching enabled, the con-

troller sends a data transfer completion

signal to the host when the controller cache

has received all data in a transaction, and

the controller then writes the cached data

to the storage device when system activity

is low or when the write buffer approaches

capacity; the cached data is not written to

the storage device immediately.

Write-through caching typically pro-

vides a data security advantage over

write-back caching, while write-back

caching typically provides a performance

advantage. The risk of using write-back

caching is that the cached data can be

lost if a power failure occurs before the

data is written to the storage device. This

risk can be mitigated by using the battery

backup units on PERC 6 controllers.

read and I/o policiesRead policies determine the type of read

option for a logical drive and can be set

to normal (the default setting), read-

ahead, or adaptive. The normal setting

specifies that the controller reads only the

requested data and does not read ahead

for the current logical drive. The read-

ahead setting specifies that the controller

reads sequentially ahead of requested data

and stores the additional data in cache

memory, anticipating that the data will be

needed soon; this policy typically supplies

sequential data faster than the normal (no-

read-ahead) setting, but is not as effective

when accessing random data. The adaptive

setting specifies that the controller begins

using read-ahead if the two most recent

disk accesses occurred in sequential sec-

tors. If all read requests are random, the

algorithm reverts to the normal setting;

however, requests are still evaluated for

possible sequential operation.

I/O policies are used to enable read

buffering in cache memory and apply to

reads on a specific logical drive. This set-

ting does not affect the read-ahead cache.

The cached I/O setting specifies that the

controller buffers reads in cache memory,

while the direct I/O setting specifies that

reads and writes are not buffered in cache

memory. This setting does not override

the read policy settings; data is trans-

ferred to cache and the host concurrently.

If the same data block is read again, it

comes from cache memory.

Software featuresEnabling disk caching can have a signifi-

cant impact on system performance, but

the potential for data loss increases if

power to the disk drives is lost while data

written to the disk is still in the cache.

Native command queuing (NCQ) for SATA

and tagged command queuing (TCQ) for

SAS hard drives helps increase perfor-

mance by allowing individual hard disks

to optimize the order in which read and

write commands are executed.

SImplIfIed managementPERC 6 controllers integrate technologies

and applications designed to simplify

management for IT administrators. For

example, SAS disk connectivity capabili-

ties have made it increasingly common to

connect a large number of disks to a PERC

controller, and PERC 6 controllers include

features that can automatically track disk

order and organization through disk roam-

ing. PERC 6 controllers also support

moving physical disks from one cable con-

nection or backplane slot to another on

the same controller; the controllers can

automatically recognize the relocated

physical disks and logically place them in

the appropriate virtual disks that are part

of the disk group.

Other advanced online capabilities

allow administrators to migrate disks from

one controller to another. The Advanced

Import feature enables administrators to

import, or migrate, disk arrays created

by a different PERC controller while the

server remains online. The controllers can

even import partial, or degraded, arrays

and provide array preview capabilities

before import to further enhance data

serviceability.

Dell PERC 6 controllers utilize Smart

Cache Technology that offers administra-

tors pre-tuned cache ratios for read and

write operations that automate cache

“As enterprise data requirements con-tinue to increase, deploying powerful, easy-to-manage RAID solutions can become essential for organizations seeking to simplify their IT environ-ments while increasing controller and storage reliability and server uptime.”


allocations based on the characteristics of

the storage environment. This auto-tuning

feature helps administrators avoid the time-

consuming effort of manually tuning the

controller cache in a test environment.

Administrators can also perform real-

time RAID-level migration by converting

one RAID level to a different RAID level.

Figure 3 lists possible RAID-level migra-

tions and the number of drives required

at the beginning and end of the process.

RAID-level migration and expansion

cannot be performed on RAID-10,

RAID-50, or RAID-60 configurations.

IncreaSed relIabIlIty and fault toleranceBeyond industry-standard RAID levels,

PERC 6 controllers offer myriad features

to help protect data, including advanced

media error protection, enhanced rebuild

capabilities, consistency check and

background initialization, and other

key features such as Self-Monitoring,

Analysis, and Reporting Technology

(SMART), failed physical disk detection,

and hot spares.

advanced media error protectionPERC 6 controllers provide advanced

media error monitoring and repair tech-

nologies designed to prevent data loss

and enable safe data retrieval. For exam-

ple, a typical response to a media error in

a RAID controller would be to read the

data from the redundant disk and con-

tinue on. However, when a PERC 6

encounters media errors during normal

read and write operations, it can immedi-

ately attempt to repair the data. This

on-the-fly media error repair capability

helps increase data reliability by fixing data

written to bad media sectors, thus helping

reduce the potential for data loss.

Preemptive media error monitoring

through disk scanning services is typically

essential for enterprise-class RAID con-

trollers. While a typical approach to this

monitoring would employ a background

consistency check to scan for media

errors, the Patrol Read feature of PERC 6

controllers implements verify commands

to help validate media—a method that can

help significantly increase efficiency, par-

ticularly when encountering multiple

media errors on drives in an array.

enhanced rebuild capabilitiesWhen an array is in degraded mode fol-

lowing a disk failure, the risk for potential

data loss increases—particularly if the

controller encounters media errors on

optimal disks in the array during a rebuild.

In contrast to controllers that might imme-

diately stop the rebuild, the PERC 6 con-

tinues the rebuild to the end, allowing

administrators to access all valid data on

the virtual disk even though data has been

lost on the virtual disk stripe where the

media error occurred.

consistency check and background initializationThe consistency check operation helps

verify the correctness of data in logical

drives that use RAID levels 1, 5, 6, 10, 50,

and 60 (RAID level 0 does not provide

data redundancy). For example, in a

system with parity, checking consistency

means computing the data on one drive

and comparing the results to the contents

of the parity drive. Dell recommends that

administrators perform a consistency

check at least once a month to help main-

tain optimum array status.

Background initialization is a consis-

tency check that is forced when creating

a logical drive, and automatically begins

five minutes after the drive has been

created. This operation checks for media

errors on physical disks and helps ensure

that striped data segments are the

same on all physical drives in an array.

Figure 3. Possible RAID migrations supported with Dell PERC 6 controllers

Source raId level target raId level required number of source physical disks

required number of target physical disks

RAID-0 RAID-1 1 2

RAID-0 RAID-5 1 3

RAID-0 RAID-6 1 4

RAID-1 RAID-0 2 1

RAID-1 RAID-5 2 3

RAID-1 RAID-6 2 4

RAID-5 RAID-0 3 3

RAID-5 RAID-6 3 4

RAID-6 RAID-0 4 4

RAID-6 RAID-5 4 3

“PERC 6 controllers are designed to provide enhanced performance, simplified man-agement, increased reliability and fault tolerance, optimized deployment with Dell PowerVault storage, and flexible RAID configuration and management tools.”

Storage


The background initialization rate is

controlled by the BGI rate set using

Dell OpenManage™ Server Administrator

(OMSA) Storage Management.

other key data protection featuresOther key data protection features of

PERC 6 controllers include the following:

SMART:■■ The SMART feature monitors

the internal performance of motors,

heads, and physical disk electronics to

help detect predictable physical disk

failures.

Failed physical disk detection:■■ The

controllers can automatically detect

and rebuild failed physical disks when

a new drive is placed in the slot where

the failed drive resided or when an

applicable hot spare is present.

Automatic rebuilds can be performed

transparently with hot spares.

Hot spares:■■ The controllers support

global hot spare, dedicated hot spare,

and affinity configurations, which

administrators can set up using the Dell

BIOS Configuration Utility as well as

OMSA Storage Management. Global hot

spares can typically be used in any

degraded RAID array when the hot

spare has sufficient capacity to fit into

the RAID container. Dedicated hot

spares are reserved for a particular disk

group. The affinity option prioritizes a

hot spare to the enclosure that it is

located in.

Transportable battery backup unit ■■

(TBBU): The TBBU is a cache memory

module with an integrated battery

pack that enables administrators to

transport the cache module with the

battery into a different controller. The

TBBU helps protect the integrity of

cached data by providing backup

power during a power outage. After

power is restored to the controller, the

cache is flushed and the protected

data is written to the disk drives.

Physical disk hot swapping: ■■

Administrators can manually substitute

a replacement unit in a disk subsystem

for a defective one while the subsystem

is performing its normal functions.

Heartbeat status LED and dirty cache ■■

LED: The heartbeat status LED on the

controllers indicates activity on the

chip, while the dirty cache LED indi-

cates that the cache has data that has

not yet been written to disk.

optImIzed deploymentS wIth dell powerVault StorageDell PowerVault storage is engineered to

work optimally with Dell PowerEdge serv-

ers and PERC controllers. Combining the

PERC 6/E with the Dell PowerVault

MD1000 disk expansion enclosure can

provide a high-performance, enterprise-

class direct attach storage solution.

Administrators can use OMSA Storage

Management software to manage both

the external array and internal storage

within the server itself, helping streamline

and simplify storage management through

a single interface while helping reduce

resource load on the system.

Administrators can customize the

PowerVault MD1000 along with the PERC

6/E to meet different enterprise needs

and help increase storage provisioning

flexibility by using one of two backplane

modes (see Figure 4):

Unified mode:■■ Unified mode allows

administrators to access all 15 hard

drives from a single controller, and

supports daisy-chaining up to two

additional enclosures per channel—

supporting a total of up to 90 hard

drives when using six enclosures and

two channels in each controller.

Split mode:■■ Split mode allows two

controllers to share the same enclo-

sure, with one having access to eight

hard drives and one having access to

the other seven hard drives. This con-

figuration does not support daisy-

chaining additional enclosures.

Unified mode Split mode

Dell PowerEdge server

Daisy-chainedDell PowerVault

MD1000 enclosures

PERC 6/E

Dell PowerEdge server Dell PowerEdge server

Dell PowerVaultMD1000 enclosure

PERC 6/E PERC 6/E

Figure 4. Unified and split modes for Dell PowerVault MD1000 enclosures with Dell PERC 6/E controllers

“Integrating PERC 6 controllers into environments using Dell PowerEdge servers, Dell PowerVault storage, and Dell Precision workstations can help administrators create a robust and highly available environment designed to maximize server uptime.”


By taking advantage of the support for

daisy-chaining in unified mode, adminis-

trators can configure a specific combina-

tion of storage resources initially, and then

add storage volumes quickly and easily as

data requirements increase.

flexIble raId confIguratIon and management toolSIn keeping with the Dell focus on simplify-

ing IT, PERC 6 controllers share a common

firmware code base, host drivers, and

management software. The storage man-

agement software enables administrators

to configure and manage multiple PERC 6

RAID systems, create and manage multi-

ple disk groups, and perform online

maintenance. By supporting multiple con-

figuration and management options

designed for flexibility and simplicity,

PERC 6 controllers can help meet wide-

ranging application and deployment

requirements.

Administrators can choose from two

primary management applications

depending on their needs: OMSA

Storage Management or the Dell BIOS

Configuration Utility. OMSA Storage

Management (see Figure 5), a storage

management application for Dell

PowerEdge servers, provides enhanced

features for configuring locally attached

RAID and non-RAID disk storage. It

enables administrators to perform con-

troller and enclosure functions—such as

creating data redundancy, assigning hot

spares, and rebuilding failed physical

disks—for supported RAID and non-RAID

controllers and enclosures from a

single graphical user interface (GUI) or

command-line interface (CLI) without

requiring use of the controller BIOS utili-

ties. The wizard-based GUI includes fea-

tures for both novice and advanced

administrators as well as detailed online

help files, while the CLI is designed for

comprehensive functionality and sup-

ports scripting.

The Dell BIOS Configuration Utility

(see Figure 6) is embedded in PERC 6

controllers and enables administrators to

configure and maintain RAID disk groups

and virtual disks and manage RAID sys-

tems. It operates independently of the

system OS. Administrators would typi-

cally use this utility for initial setup, and

then configure advanced features through

OMSA Storage Management.

SImplIfIed raId confIguratIon and managementThe Dell PERC 6 family of SAS RAID con-

trollers is designed to provide enhanced

performance, increased reliability and

fault tolerance, and simplified manage-

ment for supported Dell platforms.

Integrating PERC 6 controllers into envi-

ronments using Dell PowerEdge servers,

Dell PowerVault storage, and Dell

Precision workstations can help adminis-

trators create a robust and highly avail-

able environment designed to maximize

server uptime.

Bhanu Prakash Dixit is a senior analyst in

the Dell Enterprise Storage Group.

Sanjay Tiwari is an engineer adviser in the

Dell Enterprise Storage Group.

Kedar Vaze is an engineering manager in

the Dell Enterprise Storage Group.

Joe H. Trickey III is a storage product market-

ing consultant in the Dell Product Group.

Figure 5. Dell OpenManage Server Administrator Storage Management

Figure 6. Dell BIOS Configuration Utility

IntroducIng the dell Perc 6 FamIly oF SaS raId … · By Bhanu Prakash Dixit Sanjay Tiwari ......

Documents

Transcript of IntroducIng the dell Perc 6 FamIly oF SaS raId … · By Bhanu Prakash Dixit Sanjay Tiwari ......